JPH0720103B2 - Error recovery method in information communication system - Google Patents

Abstract

An error recovery method for use in an information communication system which comprises a plurality of nodes connected by links. Information is transferred between the nodes in frames of predefined types, including at least a first frame type used to transfer data and a second frame type used for error recovery. Each node has at least a first and a second mode of operation, in the first mode frames of both first and second types are accepted, in the second mode frames of the first type are discarded and only frames of the second type are accepted. A master node which controls error recovery is selected from amongst those nodes which can initiate transfers. <IMAGE>

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION The present invention relates to the field of information communication between nodes in a network, and more particularly to recovery from errors occurring during the transmission of data between nodes.

[0002]

BACKGROUND OF THE INVENTION Various different types of network configurations have been proposed and used for transmitting data between interconnected nodes in a network. For example, a local area network (LAN) typically comprises a number of computer-based devices distributed within a facility. LANs are often arranged in one of three basic topologies: star, bus and ring. More complex networks can be formed by interconnecting several different LANs with switches in the form of bridges or routers.

These networks provide some type of error recovery method. This includes means for detecting errors and means for correcting errors. Means for detecting errors generally include a parity check for each byte or small amount of information transmitted. Usually, the parity check is to add a single bit by the transmitter to make an easy arithmetic sum of all the bits to be transmitted, eg in a byte, making it even for even parity and odd for odd parity. . Parity checking typically detects single-bit errors and is used primarily for short-range, high integrity transmission paths.

For long distance or low integrity transmission paths,
A cyclic redundancy check (CRC) can be used. Generally, the CRC consists of one of two bytes that are sent in blocks of bytes, eg 128, 256 or 512 bytes after each. The specific CRC used is
It is defined by the polynomial used for the calculation. A general polynomial is given by the following equation.

[Equation 1] X ¹⁶ + X ¹⁵ + X ² + 1

After the data, the least significant byte of the two bytes is taken and transmitted. The receiver computes a polynomial on the incoming data using the same method and compares the computed result with the incoming CRC. The probability of detecting an even number of multi-bit errors, such as those associated with long transmission paths, is very high.

All of these error detection methods correct an error by requiring the receiving node to retransmit a block of data that contains the error. The retransmitted block is then examined in the same way as the first transmitted block. These methods do not allow the sending node to take corrective action if the receiving node does not acknowledge receipt of the data. As a result, the sending node will continue to send frames until the buffer in the receiving node is full. The drawback is that data is rejected by the receiving node until its buffer is cleared, so that further communication with the receiving node is not possible at this point to aid in error recovery. To clear its buffer, the receiving node may pass incomplete data. If the data being sent replaces the previous one,
This is another drawback, since for example an updated version of a previously stored file may corrupt the previously stored file with incomplete data.

[0007]

When there are a plurality of routes from the transmitting node to the receiving node, corrective action is taken to clear the buffer associated with the faulty route, and incomplete data is transferred from the receiving node. Rather, it can be guaranteed to be destroyed. Usually, this recovery is coordinated by one node accessing all receiving and transmitting nodes in the network. This access can be via other receiving / transmitting nodes. For a given network, this one node, the master, is permanently defined. This means that we want to define as a master a user of the network, eg a node with a different application, which is inputting or outputting information to or from one of the receiving nodes that is to be sent to another user. There is a drawback that there is. Another drawback is that the error recovery coordination of the whole network will not work if the master mode fails.

[0008]

SUMMARY OF THE INVENTION It is an object of the invention to comprise a plurality of nodes connected by links, at least a first frame type used for transferring data and a second frame used for error recovery. Information is transferred between the nodes using at least one of the links in a frame of a predetermined type including the type, information transfer is initiated by two or more nodes, and the nodes are at least the first
And an error recovery method for use in an information communication system having a second operation mode. The method defines a master node that controls error recovery upon detecting an error in a link connecting the first and second nodes, and one or both of the first and second nodes to the first node. Switching from a mode of operation to a second mode of operation, discarding frames of the first type by nodes on all of its ports and accepting frames of the second type by nodes on any of its ports; Sending a second frame type containing the first error to the master node by one or both of the one or the second node and a second error information to all nodes that can initiate the information transfer. And transmitting a second frame type including a first frame by a master node and opening a frame of the first type using a link connecting the first and second nodes. Sending, to each node to which the frame of the first type has been sent by another node that is doing so, when the sending step has been successfully completed, sending an instruction to discard the first frame type; The first and second by the master node that the type of frame is accepted.
The master node with a second frame type for canceling the second error information to all nodes capable of initiating the information transfer, by transmitting to the frame of the node of the node causing the node to switch to the first operation mode. And sending.

Preferably, the step of defining the master node preferably selects the master node according to predetermined criteria with agreement between all the nodes that can initiate the information transfer. Preferably, each node capable of initiating information transfer has a specific identifier, and the predetermined criterion is that the master node selects the node having the highest value among the specific identifiers. Is desirable.

Preferably, the error recovery method further comprises, before switching one or both of the first or second nodes to the second mode of operation, by the first or second node to the respective other nodes, Retransmitting the transmitted frame when an error is detected and checking if the retransmitted frame is received by each other node without error, if the retransmitted frame is error free If received, the error recovery method is terminated.

Preferably, the first error information identifies that an asynchronous event has occurred at one of the ports of the node to which the first error information was sent, and the second error information indicates that a link state change has occurred. It is desirable to identify Preferably, the method further comprises the step of the first node detecting the addition of the third node to a port at the first node, and the first node from the first operating mode to the second operating mode. Switching, frames of the first type are discarded by the node on all of its ports, frames of the second type are accepted by the node on any of its ports, and the first node makes a third error Transmitting a second frame type containing information to the master node, and upon successful completion of the transmitting step, transmitting the second frame type containing the unique identification of the master to the master node.
A step of transmitting by the node to a third node, and by the master node a frame to the first and third nodes which allows them to switch to the first mode of operation and to accept frames of all types. Desirably, the steps include: transmitting, by the master node, to all nodes that can initiate the information transfer, a second frame type that includes the unique identification of the third node.

Preferably, the link is a serial link.

[0013]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT Bits in an uncoded data byte are numbered 7 to 0 from left to right, with bit 7 being the most significant bit. The most significant byte of the integer is 1. Bit values are displayed as eg 1b, hexadecimal values are eg A2
Displayed as h.

The addressing scheme used in the present invention identifies three types of nodes according to their connection. These are single port nodes, dual port nodes and switches (3-16 ports). In a network using the present invention, these nodes are typically electronic devices, such as computers, printing devices, storage devices, etc.

FIG. 1 illustrates a dual port node 10 with two ports 16 and 18 connected to serial links 12 and 14, respectively.
Indicates. Also provided is a 3-way routing function 20 that connects ports 16 and 18 to node function 22. Depending on the address field, the routing function 20 sends the inbound frame to the node itself or to the outbound line on another port. When a node wants to generate a frame,
Instruct the routing function 20 to transmit the frame at the designated port. All messages and data frames for a particular command use the same port.

The following are the types of networks that can be implemented. (1) Dedicated Connection Figure 2 shows the simplest case of a dedicated connection between two single port nodes 30 and 32. (2) Strings Figure 3 shows a linear network of dual port nodes 36,38,40,42 known as strings. To allow unrestricted communication between any two nodes, the maximum number of nodes in the string is
Seventeen. This is because in the preferred embodiment, one hexadecimal digit for any other node in the string is selected for the address of one node in the string. For example, this allows you to connect 16 devices to one adapter port. The node at either end of the string can be a single port node 34, a dual port node 42 with one unconnected port or a switch. (3) Loop Loop is a dual port node 44,45,4 as shown in Fig.4.
It is a circular network containing only 6,47,48. Any 1
The loop provides better usability than the string, because the failure of one node cannot disrupt communication between any pair of the remaining nodes. Nodes can be inserted into or removed from the loop without interrupting communication between other nodes. In order to maintain these enablement characteristics, the maximum number of nodes is limited to 17 for the same reasons explained in the string network above. (4) Switch Figure 5 shows two switches 106,114 and three strings 100,1
02, 104; 108, 110; 116, 118 and a composite network including a circular path connecting a node 118 to a switch 114. Switches allow interconnection of multiple nodes. They also allow alternative routes to be provided to achieve fault tolerance.

[Frame] The two ports at the opposite ends of one serial link between nodes communicate in units called frames. A frame consists of a sequence of information bytes separated at each end by special protocol characters known as flags. The frame is divided into a sequence of 3 fields or 4 fields as shown in FIG. Control field: The control field is always present and its length is 1 byte. The control field represents the frame type. Application frames are used to transfer messages and data during normal operation. Privilege
The (privileged) frame is used for formation and error recovery. Privileged messages related to error recovery will be described later. The control frame is used for reset. Address field: The address field is always present and its length is variable between 1 and 6 bytes, depending on the complexity of the network and the number of channels realized by the destination node. This field is divided into a path component, a channel component and a pad component. The route routes the frame through the network to the destination node. A channel consists of a function within the destination node that either receives a message or receives a single data transfer. One channel for receiving the message is predefined.
All other channels are dynamically allocated for data transfer. The pad is a single number that makes the address field the total number of bytes, if necessary. The numeric value of the pad is usually unimportant, because the destination node has allocated the channel and knows the number of digits needed to address it. Data field: The data field is optional and its length can vary up to 128 bytes.
This field carries application data or messages. Outside the message defined later, the data
The contents of the fields are irrelevant to the invention. CRC field: The CRC field is always present and its length is 2 bytes. This field is the standard cyclic redundancy check for control, address and data fields. The destination shall not consider any of these fields valid until the CRC fields have been received and examined. This typically requires each port to buffer at least one frame in the receiver. Since the control and address fields may change, the routing function at the dual port or switch node does not have CRC when forwarding frames.
You have to regenerate the fields.

The maximum lengths of the address and data fields are chosen to balance network size, communication efficiency and implementation cost.

[Layer] There are two layers in a serial link, a transport layer and an upper-level protocol. The transport layer defines the following functions. Protocol, eg frame, flow control and addressing; link management, eg buffer store, port state, reset, shaping and error recovery; and physical medium, eg coding, modulation, clocking,
Line driver / receiver, connector and cable.

Each of the serial link implementations is responsible for defining: · Supported data rates, ie 10 MB / s or 20 MB / s
Or both, and-Formation of frame buffer in each port.

The associated upper level protocol defines the following functions: -Interpretation of user-defined characters. The content of the data fields in the application frame, eg commands, status and data.

The upper level protocol initiates data transfer by exchanging message frames between source and destination nodes. The destination node allocates a channel to receive the data frame and indicates the number of bytes it currently accepts.

[Response] In order to achieve the required flow control, the destination acknowledges, that is, a pair of consecutive ACK (acknowledgement) protocol characters, and the receiver available (RR = Receiver Ready), i.e. continuous
Send two responses to the source, one for the RR protocol character pair.

These protocol characters are used in pairs to protect the response from being generated by transmission errors. A node only responds to a response when it receives both characters of the pair without any intervening characters.

In the network, this protocol operates independently on each serial link. Responses are never forwarded by the routing function.

[Frame Type] The transport layer identifies the following three types of frames. -Control frame: The control frame is used for reset. The length of the data field must be 0,
Otherwise, the receiving node rejects the frame. The control frame is directly operated by the destination node. They can be sent when the transmit port is in privileged mode or normal mode. Privileged frame: The privileged frame is used by the transport layer for formation and error recovery. They can be sent when the transmit port is in privileged mode or normal mode. Application frame: The application frame is used only by the upper level protocol. The content of the data field is independent of the transport layer. When the port is in privileged mode, the transmitter discards application frames.

The content of the data field of each frame consists of commands, status or data. For each command, the node can be divided into an initiator, that is, the node that issued the command, or a target, that is, the node that received the command. Port mode: The port can operate in normal mode or privileged mode. Normal mode allows a port to send any type of frame over the link. Port has Set_normal_mode message
When (after-mentioned) is received, it shifts from the privileged mode to the normal mode. In privileged mode, the port only sends control and privilege frames to the remote node. The application frame is discarded by the transmitter. Link ER
Privileged mode is started from normal mode when there is a link error that cannot be recovered by P. Privileged mode is also initiated when the port is reset. • Disable state: The port is disabled when it is reset (for example, at power up) or when the node experiences a catastrophic internal error. In this state, all communication except reception of the reset control frame is prohibited. The port transmitter displays the disabled status by continuously sending DIS characters to the link. The port normally exits the disabled state automatically if it is reset.

[Master] Exactly one initiating function in the network is designated as the master. The application can select itself as the master. Alternatively, the master can be automatically selected by the transport layer during formation. For example, the starting function with the highest Unique_ID can be selected.

During the formation process, the master informs all other nodes of its location by issuing a privileged message to each node in turn. Each node then reports an asynchronous event by sending a privileged message to the master. (The typical event is an error that cannot be recovered by the transport layer.) The masters send privileged messages to each other to warn the initiating function and coordinate recovery actions.

If the master node fails or is disconnected, a new master must be appointed to replace it. And the new master should inform every other node of its location.

[Formation] -Unique identification: When a node is constructed, all initiating functions and switches are assigned a unique identification (Unique_ID). The unique identification is typically stored in EPROM (Erasable Programmable Read Only Memory). This is a 4-byte vendor (vendo
r) consists of an identification followed by a 4-byte node identification assigned by the manufacturer. Both identifiers are unsigned binary integers. Unique identification is used during error recovery,
Identifies the command issued by a particular start function. -Procedure: Each of the potential initiating functions must perform a formation process in order to determine the other nodes that exist and their route addresses. Configuration table: In the formation process, every initiation function builds a configuration table with entries for every other node. The entry contains a description of the node (the number of operational ports it has, and its unique identification), as well as the route address from the initiating function.

When powered up, each start function forms a whole network. When a new link is connected to the network, the initiating function must also perform an additional partial formation. In this case, a privileged message alerts each of the initiating functions, as described below.

Privilege messages also warn each of the initiating functions if there is a link error that cannot be recovered by the transport layer. If the error is permanent (eg, the link is broken), each of the initiating functions creates a route to the node beyond the scope of the error by deleting them from the configuration table. exclude. Start function table: During the formation process, every goal builds a start function table. Each table entry contains a unique identification of the initiating function and returns an address from the target to the initiating function. If the initiating function is using alternate routes to the same target, the table contains one entry for each route.

During error recovery, the start function table is used to quiesce outstanding commands.

The transport layer of the [Error Recovery] link is
It includes a link error recovery procedure (link ERP) that attempts error recovery by retransmitting the last one or two frames.

Here, a method of error recovery in a complex network having a plurality of starting functions will be described. This scheme guarantees data integrity and minimizes the impact on other operations. For example, an error prevents writing incorrect data to the disk drive. Also, the error affects only the commands, initiating functions, and targets that are currently using the failed link or node.

[Link error] -Hardware error: When the port detects an internal hardware error, for example, a parity check error, this error is displayed.・ Line fault: When the line driver or receiver detects reactive voltage and the port is not in the disabled state,
This error is displayed. In this case, the cable circuit may be open or short circuited, or the remote node may lose power. Acknowledgment (ACK) timeout: The source port is not
This error is displayed when no ACK response is received.

[Receiver error] Loss of synchronization: This is displayed when the clock recovery circuit in the receiver detects a synchronization error. Code violation: If the receiver does not detect a ^ sync loss error and it is not within a defined letter or character and decodes a character that causes a parity violation, this error is displayed. . Protocol error: If the above receiver error does not occur and the port receives the wrong sequence of valid characters listed below, this error is displayed. (1) A short frame with less than 4 data characters between two flag characters. This can be caused by dropping due to noise or flag generation. (2) Privilege frame or application frame is available but buffer is not available, i.e. RR pending (RR_pe
nding) is set. (3) Separated RR character. If no error is detected and the RR response is lost, for example if both RR characters are changed to flag characters while the link is idle, half of the link is stopped. This is unlikely, so no recovery is provided at the transport layer. instead of,
The application should give a timeout for each operation in progress. (4) Unexpected ACK response. That is, wait for positive response (Waiting_fo
r_ACK) is reset. (5) Separated ACK character. If the ACK response is dropped, the transmitter also detects the ACK timeout. (6) Null character (N
UL). (7) Abort character (ABORT) with no intervening data characters since the last flag. (8) A censoring character that is not immediately followed by a flag. CRC error: This error is indicated if the received frame has a wrong CRC, the frame is not truncated and none of the receiver errors mentioned above have occurred. Sequence error: the received frame has a frame sequence number that is not equal to the received sequence number, the frame is not truncated, the frame is not a control frame and the receiver error can occur If not, this error is displayed. The previous frame is probably lost. Frame Rejection: This error is displayed when a frame is received correctly without the above receiver error and is not truncated but the frame is unacceptable for any of the following reasons: (1) The frame contains more than 137 data characters.
Note that the receiver must continue to accumulate CRC until the trailing flag to check for transmission errors, eg, no missing flags. (2) Otherwise, the length of the frame is unacceptable for implementation, eg the message frame is too long. (3) The control field is invalid. (4) The address field specifies a destination that is not realized or is currently invalid. (5) The length of the data field in the control field is 0
is not.

Errors in this class are generally due to programming, synchronization or compatibility issues.

[Link Error Recovery Procedure (ERP)] A link error recovery procedure (ERP) is defined for a link to recover a link error at the frame level. This has the following advantages: (1) The recovery is transparent if successful, simplifying the per-level protocol. (2) Normally, no action needs to be terminated when an error occurs. However, the extra time taken on the link ERP can result in an overrun of a device with limited buffer storage. (3) The remote node has no uncertainty about the state of the application. (4) Compatibility of different link implementations is enhanced.

It is expected that the Link ERP will be implemented in firmware that normally runs on the node processor. However, the function can be imagined to be performed by a hardware finite state machine if performance is critical.

If the ERP determines that a transmission error has occurred, then E
The RP itself attempts to recover from the error. If the recovery is successful, the link ERP is terminated and the upper level protocol remains unaware of the error. ERP cannot transparently recover certain errors, such as hardware errors or permanent line failures. So that both nodes always identify the unrecoverable error and stay in sync,
ERP is carefully designed. In these cases, ERP
Escapes. If possible, recovery is attempted by command retry, as described below.

[Principle] The basic principle of the link ERP is as follows. (1) Only the broken link calls the link ERP. No other link in the network is called. (2) Link ERP is a privilege frame and application
Recover only frames. Link ERP does not recover control frames. (3) In normal operation, the transmitter does not discard the privileged frame or application frame until it receives an ACK response. This indicates that the frame was correctly received by the destination port. Thus, in the event of an error, the affected frame is still available for retransmission without reference to the upper level protocol. (4) If an error is detected, both ports will link ERP.
And exchange status by link reset. (5) Recovery is performed separately for each line. Each port is responsible for recovering lost frames on its outbound line. The sending device must be able to
Up to 2 frames may need to be retransmitted as another frame can be started to be transmitted. (6) The link ERP forcibly disables the port before restarting communication. This synchronizes the ERPs at both nodes and allows for regular restarts by the same mechanism used at power up. (7) Link protocols and ERP are designed to minimize the possibility of lost or duplicated frames when errors occur. However, upper level protocols should protect against these events whenever possible. For example, a zero byte count can be checked at the end of the data transfer and the time out can be used to detect lost messages.

[Link Status Byte] During error recovery, the link ERP in each node builds a link status byte and sends it to the other nodes in the address field of the link reset frame. The link status byte is defined as follows.

[Table 1]

H / W (hardware) error: When set to 1b, this bit indicates that the port has detected an internal hardware error. Line Fault: When set to 1b, this bit indicates that the line driver or receiver has detected a fault on the line. It is provided for diagnostic information only and is not referenced by the link ERP in the destination node. ACK T / O: When set to 1b, this bit indicates that the transmitter timed out while waiting for an ACK response.
It is provided for diagnostic information only and links within the destination node
Not referenced by ERP. Receiver Error: This field contains the following 3-bit code that identifies the first error detected at the receiver. 0 0 0 No error 0 0 1 Loss of synchronization 0 1 0 Code violation 0 1 1 Protocol error 1 0 0 CRC error 1 0 1 Sequence error 1 1 0 Frame rejection 1 1 1 Reserved

When more than one error occurs at the same time, the lowest number is reported. RSN: This is the receive sequence number of the last privileged or application frame acknowledged by the port. It is needed by the link ERP in the remote node.

[Definition of Link ERP] To facilitate cross-reference, when ERP fails, each ^ exit ^ is identified by the following name.

The first (or only) port that detects an error calls its link ERP. And the link ERP proceeds as follows. (1) ERP means that the transmitter is sending the current frame, if any,
Wait until the transmission is completed. The transmitter can arbitrarily choose to abort the current frame. (2) The ERP then builds the link status byte by referring to the hardware. (3) If the line driver or receiver detects a line fault, ERP will attempt to reset the error. If this fails, the ERP escapes and a permanent line failure occurs.
Display ^. (4) ERP checks whether the receiving device detects the DIS character. If so, the remote port may have been disabled due to a catastrophic error. ERP escapes and displays ^ remote port prohibited. (5) ERP builds a link reset containing the link status byte (see below). It then sends two consecutive link reset frames to the remote node. Such repeated link resets cause any of the frames to be dropped due to noise. The remote port should now initiate a check state, if it has not already been initiated. In any case, it calls its link ERP and returns two link resets containing the remote link status byte.

The following is the format of the link reset frame. Control field Message_code = 0CH Address field Link status byte Data field Must be absent CRC field Must be correct

Link reset is limited to a single link. It is never propagated from one link to another. (6) ERP checks if a link reset has already been received from the remote node. If not, ERP initiates a timeout and waits for a link reset. If the local node does not receive a link reset within 1 ms after sending another link reset, ERP exits and the link reset fails
Display ^. (7) If there is a permanent error, the embodiment must protect against ERP looping. The following is an example of one method that can be used.

Each invocation of ERP increments a retry counter which is periodically reset by the timer to zero. If the number of retries in one timer cycle exceeds a certain maximum value, ER
P escapes and displays ^ retry limit exceeded ^. This scheme also protects against excessive use of ERP in case of severe external noise. (8) If either port detects a hardware error, ERP exits and displays a ^ hardware error ^. (9) If either of the ports displays ^ Frame Rejection ^, further communication may be meaningless. ER
P escapes and displays a ^ frame rejection ^. (10) Otherwise, ERP calculates the number of outbound frames for which the acknowledgment is outstanding.

ERP also calculates the number of outbound frames that the local port is expecting an acknowledgment for but has not been received by the remote port.

If both of these tests fail,
ERP escapes and displays the ^ Invalid Retry Status ^. (11) Otherwise, ERP arranges to retransmit the dropped frames. (12) Outbound buffers that do not require retransmission should be discarded here. (13) If the port is receiving a frame containing one of the errors listed in ^ Receiver error ^, the appropriate inbound buffer must be discarded. Otherwise, ERP may not handle the inbound buffer. If any are full, they are emptied by the upper level protocol. (14) ERP disables the port, hardware error, A
Reset all CK timeout and receiver error latches. (15) When the receiver indicates the detection of a DIS character, the ERP waits until the remote port enters the disabled state. This is required to synchronize the two link ERPs and prevent the transmitter from sending frames while the remote port is not in normal mode.

If the receiving device does not detect the DIS character within 1 ms after the local port is disabled, the ERP exits and displays ^ Waiting timeout in ^.
The remote port may be displaying an irrecoverable error. (16) Otherwise, ERP enables the port. (17) ERP waits for a port to become available. This indicates that the remote port has finished its recovery.

1 m after the local port is ready for use
If the link is not available within s, ERP exits and displays ^ Availability timeout wait ^.
This may indicate that the remote node has been powered down or has suffered a catastrophic error. (18) Otherwise, ERP ends well.

[Non-Link ERP Error Recovery] The action to be taken by each node when the link ERP fails and exits is the link
It is described as well as a set of primitive and some procedures for handling events outside the scope of ERP. These events are as follows: (A) Link ERP fails. This may be a transient unrecoverable error or a permanent error such as a broken link. (B) Target node does not respond to commands. (C) The routing function receives a frame addressed to a non-operational port. (D) Node receives invalid message but does not know return address. (E) A new link is connected to the network.

For the first four events (A-D), the initiating function generally terminates the affected commands and retries them using an alternate path if necessary. Last case
In (E), none of the commands are affected.

Privileged Message Description The following section defines messages that support error recovery. In all cases, the control field represents the frame type of ^ privilege. -Set Master (Set_master): This message is defined as shown in Table 3 and is a master (Master).
From every other node in the network during the formation process. This is a node alert to master linked
(Link_alert) Specify the return address (Return_address) and tag (Tag) used when sending a message. The destination node records this information and sends back a response message.

[Table 2]

Message Code: This byte identifies the message as the Set Master. • Tag: This 2-byte field is returned in the response message. The same tag is also used in the link warning message if the node then reports a link state change. The tag is specified by the master and must be the only among the currently active tags from the master. The tag remains active until the node receives another set master message. Return Address: This 4-byte field specifies the value to be placed in the address field of the resulting response message followed by the link warning message. If a node receives a set master message and has a node in privileged mode, then the node should send a link warning message to the master before sending the set master message response. The node only stores the return address and tag of the latest set master message it received.

Response: This message is shown in Table 4
Defined as set master, master alert,
Quiesce and set normal mode (Set_normal_mod
e) Bounced to acknowledge the message. A reply is sent to the same port that received the source message.

[Table 3]

Message Code: This byte identifies the message as a response. • Return Code: This byte is set to 00h if the source message was processed successfully. Any other value indicates that the requested function could not be completed. • Tag: This 2-byte field is copied from the source message. It identifies the message that is about to be acknowledged. The address field in the reply message is obtained from the return address field in the message about to be acknowledged.

Link warning: This message is defined as in Table 5. The node sends this message to notify the master of an asynchronous event on one of its ports. The master does not respond directly.

[Table 4]

Message Code: This byte identifies the message as a link warning. Type: This type is encoded to represent the event. The main event types are as follows. The port is currently operational. This means that the new link is connected. • The addressed port is not ready. This indicates that the routing function received the frame but could not forward it. -Message rejection. This is because the node received an invalid message but does not know the return address, for example the message
Indicates that the code is invalid. -Permanent disability. This means that the link ERP will not work, eg because the link is broken. -Unrecoverable error. This represents a temporary error that could not be recovered by the link ERP. -Remote port does not respond. This indicates that the remote port did not respond during the link ERP. -The remote port is prohibited. Link ERP started because port received DIS character. For example, the remote node may have been reset or suffered a catastrophic internal error. Tag: This 2-byte field contains the tag specified by the latest set master message. It allows the master to determine which node sent the link alert. • Port: Bits 3: 0 contain an unsigned integer identifying the affected port. The address message in the link warning message is obtained from the return address field in the latest set master message.

Master Warning: This message is defined as shown in Table 6 and is sent by the master to each other initiating function. It has two uses: (1) Transfer a link warning for an asynchronous event. (2) Indicates that both ports of the link have been moved to normal mode following error recovery, for example. In either case, the initiating function sends a response message back to the master.

[Table 5]

Message Code: This byte identifies the message as a master alert. • Type: If the master alert is forwarding a link alert, this byte is copied from the corresponding link alert. In this case, the destination initiation function should quiesce any command that was using the specified route before returning the response. If the master indicates that both ports of the link are returning to normal mode, this byte will
It is set to FFh. Tag: This is a 2-byte field assigned by the master and returned in the response from the destination initiation function.
It must be unique among all currently active tags from the master. Return Address: This 4-byte field specifies the value to be placed in the address field of the response message. • Route: This 4-byte field specifies the address of the node that generated the link alert for the initiating function that received the master alert. Port: Bits 0: 3 are copied by the master from the corresponding link warning message.

Quiesce: This message is defined as shown in Table 7 and is sent from the initiating function to the target during error recovery to quiesce all commands from the specified initiating function. The target sends a reply message back after it quiesces the affected command. The target does not send back the status of quiesced commands.

[Table 6]

Message Code: This byte identifies the message as stationary. Tag: This is a 2-byte field assigned by the start function and returned in the response from the target. Return Address: This 4-byte field specifies the value to be placed in the address field of the response message. Unique ID: This is the 8-byte unique ID (Unique_ID) of the start function where the command should be quiesced. Before selecting a command to quiesce, the target must search its initiating function table and translate the unique identifier into the return address. Specifying the start function by a unique identifier rather than the return address allows the start function to use an alternate route for quiesce if the first route is no longer available. It allows the master to issue a third party static on behalf of the missing start function.

Set Normal Mode: This message is defined as shown in Table 8 and is sent by the master to change the port from privileged mode to normal mode. The destination mode sends back a response message. The link can be used for application frames when the ports are changed to normal mode on both ends of the link.

[Table 7]

Message Code: This byte identifies the message as set normal mode. • Port: Bits 3: 0 contain an unsigned integer that identifies the port that is changed to normal mode. Tag: This is a 2-byte field assigned by the master and returned in the response from the destination node. It must be unique among all currently active tags from the master. Return Address: This 4-byte field specifies the value to be placed in the address field of the response message.

Recovery Procedure This section shows how to use the previous concepts and privileged messages during error recovery.

(A) Link ERP fails This means that the error could not be recovered by the transport layer. Both ports of the failed link go into privileged mode, so their transmitter discards the application frame. This avoids frames that hack up and block other traffic. This also prevents further data from being transferred by the affected link and written to the magnetic media.

If the return path specified by the set master during formation contains a failed link, each node sends a link warning message to the master. Therefore, the master receives one or two link warning messages.

Then, the recovery proceeds as follows. (1) If the link warning indicates ^ remote port is not responding ^, the master immediately issues a reset control frame to the failed node.

The master then unforms the affected links and nodes. (If the failed node recovers after a reset, the adjacent node will generate another link alert indicating that the ^ port is now operational ^, which will reshape the affected link and node.) (2) If Link warning ^ Remote port disabled ^
If so, the remote node may have been reset or suffered a catastrophic internal error. In the first case, the remote node normally re-enables the port itself. In the second case, it may be necessary to reset the node. The master waits up to 1 second to receive another link alert from its neighbor indicating that the port is currently operational. This link alert indicates that the node is currently recovering. If no link warning is received, the master assumes the cause was a catastrophic error and issues a reset control frame to the node. In this case, the master unforms the affected links and nodes. (3) If the link warning displays ^ Permanent failure ^,
The master also unforms the affected links and nodes. (4) The master issues a master alert to each of the other initiating functions that remain in its formation table. (5) If the master alert indicates a permanent failure or the node has been reset, other initiation functions will de-form affected links and nodes. (6) Each initiating function identifies the command that was in progress on the failed link and stops the associated outbound data transfer. (7) If the targets remain formed (perhaps due to the use of alternate routes), each initiating function issues a quiesce message to each target that was executing the affected command.

If the master determines that an error has removed the previous initiation function from the network, it will issue a third party quiesce to all remaining targets on behalf of the missing initiation function.

The target sends back a response message for each quiesce. When all of the affected targets are quiesced, each other initiation function sends a master alert message response message to the master. (8) If the link is still formed, the master sends set normal mode to each port. The master waits for a response message in set normal mode. (9) The master sends a master warning message to each of the other initiating functions to indicate that both ports are currently in normal mode. Each initiating function sends back a response. (10) If the link is still formed or if there is an alternate route, each initiating function reissues the affected command.

(B) The target does not respond The start function is an anti-hang (anti-ha) command for each command issued to the target.
ng) Expected to start a timer. The timer is stopped when the start function receives a situation indicating that the target has finished processing the command. The timer is protected against undetected link failure on the path to the target node or software failure on the target itself. If the timer expires before it is stopped, the start function should operate as follows. (1) The start function issues a privileged message to sequentially ask each intermediate node that starts at an adjacent node. If any of the modes do not respond then the node is assumed to be hung. Otherwise, the target node is hung. (2) The start function issues a reset control frame to the hung node. This will disable all ports on the node. (3) The port connected to the adjacent node detects this, calls the link ERP, and generates a link warning indicating ^ remote port disabled ^. This is processed as described above.

(C) Addressed port does not work Frames may be incorrectly addressed to non-working routing function ports. This is processed as follows. (1) The node detecting the error sends a link warning to the master, indicating that the addressed port is not working. The port remains in privileged mode. (2) The master issues a master alert to each of the other initiating functions and specifies that the addressed port does not work. (3) Each initiating function terminates the affected command and associated outbound data transfer and returns a response. (4) If there is an alternate route, each initiating function reissues the affected command.

(D) Invalid message received If the destination node knows the return address (eg, the message has invalid parameters), it returns a response with a non-zero return code. Otherwise, the destination node will generate a link warning specifying ^ message rejection ^. This means that the ^ addressed port will not work unless the port stays in normal mode ^
Treated as if.

(D) When the new link is connected to the network to which the new link is connected, the port on one of the existing nodes becomes operational. (1) The node sends a link warning to the master specifying that the ^ port is currently operational ^. (2) The master forms a new node. (3) The master issues the set master to the new node. Each node sends back a response. (4) The master issues a link warning and other new ports to the port that created it and puts it in normal mode. Each port sends back a response. (5) The master sends each other initiation function a master alert that specifies that the ^ port is currently operational ^. (6) Any other start function forms a new node.

[Embodiment] The error recovery method described above can be used in various different applications. Two examples will be described below. It is understood that the present invention can be readily used with other types of networks. As shown in FIG. 8, the dual port device string is particularly attractive for connecting I / O devices to personal computers. The adapter 50, which generally exists in the system unit of a personal computer, is a link 51.
The disk device 52 is connected to the disk device 52 via the link 53, the disk device 52 is connected to the disk device 54 via the link 53, and the disk device 54 is connected to the printing device 56 via the link 55. The use of strings reduces the connection cost per device and avoids wiring congestion at the adapter. The use of the error recovery technique of the present invention prevents errors that result in incorrect data being written to or printed on the disk drive. The optional provision of link 57 allows the loop to be closed, which provides greater bandwidth or constant fault tolerance. Since the error only affects the command using the failed link or node, the closed loop still allows access to all nodes except the failed node.

FIG. 9 shows a typical network configuration that can be used as a high availability file server.
High availability is important in such shared systems. This application also requires a dual port disk device, but here the main reason for requiring a second port is to provide a backup path in case of a failure on the main connection path. Therefore, in practice, all serial disk devices probably have dual ports. With respect to the disk array, as shown in FIG. 9, the dual port disk drive allows for a single point of failure configuration.
In this configuration, a pair of servers 60 and 62 are connected to both switches 64 and 66 via dedicated links, and each port of dual ported disk devices 68, 70, 72 and 74 is connected to one of the switches. . The use of a dedicated link to each disk unit allows fully concurrent maintenance without affecting the operation of other disk units.

[Brief description of drawings]

FIG. 1 is a diagram of the main functional component of a dual port node as may be found in a network using the techniques of the present invention.

FIG. 2 is a block diagram of a simple network including a pair of interconnected single-port nodes.

FIG. 3 is a block diagram of a network including a string of interconnected single-port and dual-port nodes having a single transmission path between the nodes.

FIG. 4 several interconnected in a loop configuration with two possible transmission paths between nodes, as shown in FIG.
FIG. 3 is a block diagram of a network including dual port nodes.

FIG. 5 is a block diagram of a composite network of interconnected nodes including single-port, dual-port and switch nodes.

FIG. 6 is a diagram showing a format of a single frame used for communication between nodes as shown in FIGS. 2 to 5;

7 is a diagram showing the format of the address field component of FIG. 6;

FIG. 8 is a block diagram of a personal computer system in which the present invention may be used.

FIG. 9 is a block diagram of a file server system in which the present invention may be used.

[Explanation of symbols]

 10 Dual Port Node 12 Serial Link 14 Serial Link 16 Port 18 Port 20 3 Way Routing Function 22 Node Function 30 Single Port Node 32 Single Port Node 34 Single Port Node 36 Dual Port Node 38 Dual port node 40 Dual port node 42 Dual port node 44 Dual port node 45 Dual port node 46 Dual port node 47 Dual port node 48 Dual port node 50 Adapter 52 Disk unit 54 Disk unit 56 Printer 57 Link 60 Server 62 Server 64 Switch 66 Switch 68 Disk unit 70 Disk unit 72 Disk unit 74 Disk unit 100 Single port node 102 Dual port node 104 Dual port node 106 Switch 108 dual-port node 110 dual-port node 114 switch 116 dual-port node 118 dual-port node Node

 ─────────────────────────────────────────────────── ————————————————————————————————————————————————— Inventor Reginald Beer 5 S.E., England 1 P.X., Chandler's Close 4 Chandlers Ford, Hampshire 4.

Claims (4)

[Claims] 1. A method of recovering a communication error in an information communication system including a plurality of nodes connected by links, wherein the system initiates the information communication by one or more of the nodes and transfers data. In a system in which information is communicated between the nodes in the form of one type of frame and a second type of frame for error recovery, the node having at least a first mode of operation and a second mode of operation, An error recovery method comprising: a) defining a master mode for controlling error recovery; and b) detecting an error in a link connecting a first node and a second node, wherein the first node or the Switching at least one of the second nodes from the first operating mode to the second operating mode, the frame of the first type being the node Discarding at all ports and accepting the second type of frame at any of the nodes; a first node from at least one of the first node or the second node to the master node; Transferring a second type of frame containing error information, and transferring a second type of frame containing second error information to all nodes that can initiate information transfer from the master node A step of using the link between the first node and the second node to initiate transfer of a frame of the first type by another node to which the frame of the first type has been transferred. Transferring a command to each node to discard the first node, and c) each step of b) On successful completion, the master node switches these nodes to the first mode of operation for the first node and the second node to accept frames of all types. And a step of canceling the error information by transferring a second type frame to all nodes capable of starting information transfer by the master node. 2. The step of defining the master node comprises:
The method of claim 1 including selecting from all nodes that can initiate information transfer based on predetermined criteria. 3. The method of claim 2, wherein the node capable of initiating the information transfer has a unique identification, and the predetermined criterion is to select the one with the largest identifier. 4. The first node is changed from the first operation mode to the second operation mode when it detects that the third node is added to the port of the first node. By discarding the frame of the first type at all ports of the first node and accepting the frame of the second type; and from the first node to the master node. In contrast to the step of transferring a second type frame including the first error information, a second type frame including the unique identification of the master node from the master node to the third node A step of transferring, and e) when the steps of step d) are completed successfully, the master node to the first node Accepting frames of all types by transferring frames to the third node for switching these nodes to the first mode of operation, and all that can initiate information transfer from the master node. Forwarding the second type of frame containing the unique identifier of the third node to the node of claim 3, the method of claim 1, 2 or 3 further.